We need a structured description of the process we use to extract relevant content from the raw transcript. I'll work on this one